O3199: Bringing Sepsis to Its Hands and Knees! Improving Early Sepsis Detection in a Pediatric Hospital Inpatient Setting

Purpose/Objectives: The Improving pediatric sepsis collaboration identifies screening tools as important for sepsis bundle compliance shown to decrease mortality. Our quaternary urban children’s hospital currently lacks a uniform sepsis screening process in the inpatient units. This project aims to improve early sepsis detection through the use of a sepsis crawler tool within the Electronic Health Record (EHR) on a pediatric inpatient unit.

Design/Methods: Our multidisciplinary team worked with an EHR analyst to program a crawler algorithm based on SIRS criteria and systemic chart data including cardiovascular data, neurologic data (GCS), renal data (increase in CR), hematologic data (ANC, high INR, drop in platelet count), respiratory data, and hepatic data. (FIg 1&2). A positive sepsis trigger results in a task to the bedside nurse to perform a prescreen for sepsis with a scoring algorithm. A negative screen will suppress the notification and a positive screen will trigger the nurse to contact the provider for a bedside sepsis huddle. There is also a trigger for severe sepsis which notifies both nurse and provider. If a sepsis huddle is triggered, provider will document the huddle process and any actions taken. Alert suppression lasts for 24 hours. All chart data/documentation from alerts is collected and reviewed. Outcome measures include percentage of true positives, time to fluids and antibiotics in positive sepsis huddle. Process measures include nursing secondary screen compliance and huddle completion. Balance measures include the rate of false positives. The crawler is deployed on a single inpatient unit (16 beds) with a gradual rollout plan over the next 6 months. A live dashboard built allows for tracking of metrics during implementation.

Results: Prior to implementation, 3 months of data were reviewed to evaluate accuracy of the EMR algorithm and used to create buy in to key stake holders. Time of development was approximately 6-8 months and required approximately 2 months of analyst time from the EHR vendor.
Screening for sepsis increased from 0 to 100% after EHR tool was implemented on the initial inpatient unit. The sepsis alert fired 22 in 4 weeks and 13 events in 4 weeks fired for severe sepsis. Of alerts, 14% had positive nursing screen escalating to physician huddle. Preliminary results demonstrate the tool is appropriately selective with the current algorithm. Current data is still being collected on time to fluid and antibiotics in patients with positive sepsis alert.

Conclusion/Discussion: The sepsis crawler in its early deployment has been an effective tool improving early detection of sepsis and there is continued plan for expansion over 6 months throughout the other inpatient units. The biggest barriers to deployment included lack of ‘EHR crawler specialist’ onsite, identification and coordination with initial inpatient site, education, and building of report to feed a real-time dashboard.